king - human hippocampus and spatial memory

Human Hippocampus and Viewpoint Dependencein Spatial Memory

John A. King,1* Neil Burgess,1 Tom Hartley,1

Faraneh Vargha-Khadem,2 and John O’Keefe1

1Institute of Cognitive Neuroscience and Department ofAnatomy, University College, London, United Kingdom2Developmental Cognitive Neuroscience Unit,Institute of Child Health, University College,London, United Kingdom

ABSTRACT: Virtual reality was used to sequentially present objectswithin a town square and to test recognition of object locations from thesame viewpoint as presentation, or from a shifted viewpoint. A develop-mental amnesic case with focal bilateral hippocampal pathology showeda massive additional impairment when tested from the shifted viewpointcompared with a mild, list length-dependent, impairment when testedfrom the same viewpoint. While the same-view condition could be solvedby visual pattern matching, the shifted-view condition requires a view-point independent representation or an equivalent mechanism for trans-lating or rotating viewpoints in memory. The latter mechanism was indi-cated by control subjects’ response latencies in the shifted-viewcondition, although the amnesic case is not impaired in tests of mentalrotation of single objects. These results show that the human hippocam-pus supports viewpoint independence in spatial memory, and suggest thatit does so by providing a mechanism for viewpoint manipulation inmemory. In addition, they suggest an extremely sensitive test for humanhippocampal damage, and hint at the nature of the hipppocampal role inepisodic recollection. Hippocampus 2002;12:811–820.© 2002 Wiley-Liss, Inc.

KEY WORDS: developmental amnesia; hippocampal lesion; virtual re-ality; cognitive map; recognition memory; allocentric; mental rotation

INTRODUCTION

Developmental amnesics with focal bilateral hippocampal damage tend tohave spared performance in recognition memory paradigms but stronglyimpaired performance in the recollection of episodic information (Vargha-Khadem et al., 1997; Baddeley et al., 2001). The profound deficit in therecollection of episodic memory also invariably results in cases of similardamage acquired in adults, while recognition memory appears to be sparedin some cases (e.g., Holdstock et al., 2000a), but not in others (e.g., Manns

and Squire, 1999; see Spiers et al., 2001b, for a compre-hensive review). What does the human hippocampus dothat is so crucial to episodic recollection, but not to rec-ognition?

One line of attack is to examine the restricted deficit ofthe developmental amnesics (Vargha-Khadem et al.,1997) in more detail. Interestingly, these cases were im-paired on two types of recognition memory: memory ofobject locations and memory for voice–face associations.In this study, we attempt to characterize the basic func-tional deficit of one of these patients, “Jon,” within therecognition paradigm and the spatial modality. Jon hasfocal bilateral hippocampal pathology with apparentsparing of other mesial temporal lobe structures (Gadianet al., 2000) (Fig. 1). We hope that isolating specificfunctional impairments in the spatial domain will pro-vide a link to the general deficit in episodic recollection.

The cognitive map theory relates episodic memory tothe human hippocampus by suggesting that it stores thespatiotemporal context of personally experienced events(O’Keefe and Nadel, 1978). This supported the thententative suggestion that episodic memory (Tulving,1983) relies on the hippocampus (Kinsbourne andWood, 1975) and was proposed as an extension of thehippocampal role in the rat, of enabling locations in theenvironment to be remembered. A cognitive map can bedefined in terms of the spatial behaviors it allows. Prin-cipally, these concern memory for locations defined rel-ative to the environment as opposed to locations that canbe approached by sensory guidance or as a series of bodymovements from a given starting point. Thus, a cognitivemap enables behaviors such as finding an unmarked goallocation from a novel starting position and taking a novelshort cut. In addition, the human Cognitive Map pro-vides the framework within which a person’s subjectiveviewpoint can be moved to different locations in an en-vironment (O’Keefe, 1993).

A strong prediction of the cognitive map theory there-fore is that patients with selective damage to the hip-pocampus, either bilaterally or unilaterally on the right,

Grant sponsor: MRC (Great Britain).*Correspondence to: Dr. John A King, Institute of Cognitive Neuroscienceand Department of Anatomy, University College London, 17 QueenSquare, London WC1N 3AR, UK. E-mail: [email protected] for publication 18 April 2002DOI 10.1002/hipo.10070Published online 00 Month 2002 in Wiley InterScience (www.interscience.wiley.com).

HIPPOCAMPUS 12:811–820 (2002)

© 2002 WILEY-LISS, INC.

will have a specific deficit in memory for locations defined relativeto the environment as opposed to locations defined by their sen-sory characteristics or by their location relative to the body. Werefer to this former type of memory as “allocentric” memory and tothe latter types of memory as forms of “egocentric” or “sensory”memory. To distinguish between “allocentric” behavior, directedtoward a location in the environment, and “egocentric” or “senso-ry” behavior, directed toward locations defined by their appearanceor position relative to the body at encoding, it is necessary for thesubject to move with respect to the environment between encodingand retrieval.

Experiments involving movement of the subject or manipula-tion of their viewpoint have played an important role in under-standing the mechanisms supporting human spatial memory.First, automatic processes may support the accommodation of achange of viewpoint within one’s mental representation of objectlocations. Wang and Simons (1999) showed that subject’s recog-nition memory for an array of objects on a circular table top wasbetter after the subject had moved around the table to a new view-point than after an equivalent rotation of the tabletop. This was

interpreted as evidence for an automatic updating process drivenby the active motion of the subject. A replication of thisresult using purely visual virtual reality (Christou and Bulthoff,1999) indicates that the important variable is movement of view-point relative to the subject’s cognitive model of the world ratherthan actual movement and the attendant vestibular/proprioceptivefeedback or movement of the objects relative to the subject. Sec-ond, when subjects are shown a scene of scattered objects from oneviewpoint and given a recognition memory test using scenes froma second viewpoint, their response latencies vary linearly with theangular difference between the views (Diwadkar and McNamara.,1997).

The above results in which subjects or their viewpoints aremoved between presentation and retrieval are also found in studieson the effects of imagined movement. Testing memory for spatiallocations following imagined movement shows a good chronomet-ric relationship between reorientation distance (either translationor rotation; Easton and Sholl, 1995) and reaction times. Increasedresponse accuracies have also been reported in memory for loca-tions from new imagined points of view that are aligned withenvironmental landmarks (Shelton and McNamara, 2001; Mc-Namara et al., 2002; Mou and McNamara, 2002). Note that noneof these results is simply analogous to the type of mental rotationfound in recognition of single objects (Shepherd and Metzler,1971). Indeed, paralleling the results of Wang and Simons foractual movement, memory for the locations of objects in an array issuperior following imagined movement of the viewer than follow-ing an equivalent imagined movement of the array (Wraga et al.,2000). Only when a single location from the array needed to berotated did performance or latencies for imagined array-rotationapproach that for imagined movement of the viewer. Thus we seeagain that movement of viewpoint relative to the subject’s cogni-tive model of the world is privileged relative to movement of theobjects within it. Note that these results hold whether or not theactual world is visible: subjects performing imagined movementsare typically blindfolded, and Simons and Wang (1998) found thesame advantage for view-point movement when subjects weretested in a dark room, using phosphorescent objects.

In the present study, we use a virtual environment (Figs. 2, 3) totest memory for object location in two conditions, one involvingmovement of the subject between encoding and retrieval and theother not. On each trial a number of objects were presented se-quentially in different locations within an enclosed courtyard whilethe subject watched from a given vantage point. Subsequently, thesubject was required to recognize the objects’ locations whenviewed either from the same viewpoint or from a different, shifted,viewpoint (Fig. 3). The two retrieval conditions were interleavedacross trials in random order. In the same-view condition the ob-ject location could be remembered relative to the virtual environ-ment (allocentric memory), or relative to the subject, the frame ofthe video display unit, or the perceptual features of the display asseen from that view (all forms of egocentric or iconic memory). Inthe shifted-view condition, using any of the latter forms of memorywould leave the subject severely disadvantaged compared with us-ing allocentric memory.

FIGURE 1. Hippocampal cross-sectional area as a function ofslice position, sectioned posterior to anterior. Upper: slices from acontrol subject (left) and Jon (right). Lower: solid lines show Jon’sdata (right hippocampus dark-shaded), the broken lines are �2 SDthe mean hippocampal cross-sectional area of a group of 22 normalhealthy subjects (Van Paesschen et al., 1997). Volumes are uncor-rected for intracranial volume. Adapted from Spiers et al. (2001a).

812 KING ET AL.

We reasoned that Jon should have preserved egocentric or iconicmemory performance but impaired allocentric memory. The pre-served egocentric memory should support some ability to recog-nize object locations in the same-view condition, but not in theshifted-view (allocentric) condition. In addition, the use of virtualreality makes recording response latencies straightforward, to allowsome insight into the processes supporting these behaviors.

MATERIALS AND METHODS

Subjects

Jon shows bilateral hippocampal pathology due to perinatal an-oxia (Gadian et al., 2000; Vargha-Khadem et al., 1997). Neuro-pathological assessment of Jon was done using MRI volumetricmeasurement (Van Paesschen et al., 1997), T2 relaxometry andproton magnetic resonance spectroscopy (1H MRS) (Vargha-Khadem et al., 1997; Gadian et al., 2000). Volumetric assessmentfound the hippocampi to be reduced in volume by 50% along thelength of both (Fig. 1). Relaxometry showed elevated hippocampalwater volumes bilaterally, suggesting disruption of remaining tis-sue. However, there are also indications from functional MRI thatthis tissue may be functional (Maguire et al., 2001). 1H MRSexamined a 2 � 2 � 2 cm area including hippocampal and sur-rounding tissue, and found neural cell to other cell intensity signals

were normal on the left and near normal on the right, suggestingthat tissues around the hippocampi are uncompromised. A morerecent analysis of MRI data using voxel-based morphometry of Jonand four other patients who had suffered perinatal or infantilehypoxic-ischemic incidents (Gadian et al., 2000) confirmed that,within the temporal lobes, the damage is confined to the hip-pocampi. Outside the mediotemporal lobes in this group there wasalso reduced gray matter density in the putamen and ventral thal-amus bilaterally.

Jon was aged 22 at time of testing. His IQ was assessed with theWAIS-R test at age 19 (full-scale IQ 114; performance IQ 120;verbal IQ 108) and performance IQ tested at age 22 using RavensAdvanced Matrices Set I (90th percentile). He regularly plays videogames using first-person perspective virtual reality. Two groups of12 male control subjects were recruited for each of the two behav-ioral experiments, and a group of 24 male control subjects wererecruited for the response latency study (Fig. 6b). All groups werematched for age and performance IQ (PIQ, estimated using theNART and the short form of Ravens Advanced Matrices Set I). Forthe behavioral data shown in Figure 6a, the values were as follows:mean age: 21.8, range 19–26 years; NART-PIQ: mean 113.8, SD7.1; Ravens PIQ: mean: 76.7th percentile, SD 18.3%. For thebehavioral data shown in Figures 4, 5a, and 5b, the values were asfollows: mean age: 20.9, range 20–22 years; NART-PIQ mean109.6, SD. 12.6 Ravens PIQ: 86th percentile, SD 8.0%. For theresponse latency data in Figure 6b, the values were as follows:NART-PIQ mean 110.8, SD 5.7 Ravens PIQ: 78th percentile, SD

FIGURE 2. Plan view of the virtual town square. Shaded areas represent places the subject wasable to traverse. The three viewpoints are marked (1, 2, and 3; with 2 used only during testing, notat presentation), with arrows showing the direction of view. Small squares mark the positions ofplaceholders on which objects were presented.

_____________________________________________ VIEWPOINT DEPENDENCE IN SPATIAL MEMORY 813

22.1%. Informed consent for Jon’s participation was obtained inaccordance with guidelines set by the Great Ormond Street Hos-pital for Children and the Institute of Child Health Ethics Com-mittee. All control subjects gave their informed consent in accor-dance with the UCL/UCLH Ethics Committee.

Computerized Tests

The virtual environment (VE) in this study was implementedusing a modified version of the computer game Quake2 (© IdSoftware). The modifications allowed timed presentation of ob-

jects and repositioning of the subject’s viewpoint. The VE waspresented in first-person perspective using an Intel P3 600-MHzcomputer on a standard 19 inch monitor at a resolution of 800 �600 pixels and a frame rate of 40 Hz. The environment consisted ofa courtyard surrounded on all sides by visually distinct buildingsand can be moved smoothly through using the cursor keys onkeyboard. Subjects were able to move around the edge of the court-yard at rooftop level, providing a clear view into it and the objectsit contained. Inside the courtyard were 21 randomly distributedplaceholders, upon which the test stimuli appeared. Subjects were

FIGURE 3. Views of the virtual environment during presentation of a typical item from loca-tion 3 (i) and testing from the same view (ii, location 3) and an alternative view (iii, location 1). Notethe red and white marker in panel (iii) showing the original viewing location during the testingphase.

814 KING ET AL.

initially given a familiarization task in the VE, in which theymoved around all of the available viewing areas. Three particularlocations were used for presentation and testing purposes, chosensuch that moving from one to the other involved a rotation ofeither 140°, 85°, or 55° in viewing orientation. The correspondingchanges in angular position of viewing location about the centroidof the placeholders were 151°, 85° and 66° (the slightly different

sets of angles occur because views were not centered on the cen-troid of the placeholders due to the necessity of ensuring that allplaceholders were visible). See Figures 2 and 3 for further details.

Testing took the following form. Subjects always appeared inthe same spot in the VE. One of the presentation positions wasmarked, and the subject moved to that marker. On contact withthe marker, their view was adjusted to a standard orientation from

FIGURE 4. Performance in the same view condition is equivalentto performance in two-dimensional object location tasks. Upper: Thetwo-dimensional object location task, shown in the testing phase. Oneof the pictures (of a saw) is in the same location as originally pre-sented, the other two pictures (foils) are in randomly chosen loca-tions. The shaded rectangles are placeholders, three of which areobscured by the objects. Lower: Performance in 2D and 3D. In both

2D and 3D test, ten objects were presented and testing involvedforced-choice recognition with 2 foils. In the 3D test, the same viewwas used for presentation and testing. Control subjects perform nearceiling on both conditions (2-D: mean score � 97%, sd � 5%; 3-D:94%, sd � 6%), while Jon is equally impaired on both (2-D: meanscore � 68%, z � 5.7; 3-D: 65%, z � 4.8).


which all placeholders were visible, and presentation began. Anumber of images of everyday objects were presented in turn for3s each, with an inter stimulus interval of 1s. Each object appearedover a particular randomly chosen placeholder, each of which wasused only once per presentation. The subjects were instructed toname the objects aloud as they appeared, and to remember theirlocations (Fig. 3). After the presentation phase, memory could beprobed from the same viewpoint or from a shifted viewpoint. Inshifted-view trials, the subject’s viewpoint was changed instanta-neously to the new location immediately after the last object hadbeen displayed. After a 5-s pause, during which the place markerswere visible in the square, testing began. Memory for the locationsof the objects was tested in random order: N copies (where N �3–6) of the object to be tested were presented, one at its originallocation and N-1 at foil locations (Fig. 3, N � 3). Each object hada different colored square superimposed on it, and responses weremade by pressing a colored key on the keyboard corresponding to

the color on the chosen object. Choices and response latencies werelogged. Both list length, as well as the number of foils, could bevaried to control the difficulty of the task. Subjects were giventraining, prior to testing, during which they practiced both same-view and shifted-view trials with list-length 5.

The 2-D task was implemented using Microsoft Visual Basic onthe same hardware as the VR task. Subjects were presented with 10pictures of objects one at a time, for 3s each with an ISI of 1s.These could appear on any of 21 randomly scattered placeholders,which were not reused as a target location during that trial. After 10objects had been presented there was a 5s pause, then the testingphase began. Objects were presented individually in a randomorder in their original location, along with two identical foils onrandomly chosen placeholders. Subjects indicated their choice,and the experimenter triggered presentation of the next object. Theprocedure was repeated four times, for a total of 40 object-loca-tions probed.

FIGURE 5. a: Performance as a function of list length for Jon andcontrols, testing controls with five foils and Jon with two foils, ex-pressed as a percentage of the range between chance (0) and perfect(100), i.e., percentage correct (x) scaled by the level of chance (c) usingthe formula: performance � 100 (x � c)/(1 � c). This allows a clearercomparison between Jon and controls, for whom the level of chance is

different (33% and 16.7%, respectively). Note that Jon’s performancefalls to chance for two objects in the shifted-view condition, but re-mains above chance for at least 13 objects in the same-view condition.b: When compared directly with controls at list length 4 and 7 (dashedlines), Jon’s performance on (solid lines) same and shifted views isclearly dissociated. Scores are percentage correct scaled as in (a)

816 KING ET AL.

RESULTS AND DISCUSSION

We first attempted to replicate the previously reported 2-D object-location deficit (Vargha-Khadem et al., 1997), and determine howthis relates to the same-view condition of our 3-D VR test. Our rep-

lication used 10 objects sequentially presented at locations selectedfrom 21 different possible locations on the screen. Testing involvedchoosing the object in the same location as at presentation in thepresence of two foils (copies of the object in other locations) (Fig. 4). Inline with (Vargha-Khadem et al., 1997), we found a mild impairmentin this task, and found identical performance of both Jon and controlsubjects in the same-view condition of our virtual town square test runwith the same number of objects, locations, foils, and timing.

We next compared the same-view and shifted-view conditions.Because pilot studies indicated that the same-view condition waseasier, we attempted to match difficulty across conditions by using10 items in same-view trials, versus 7 items for shifted-view trials.In the testing phase, two foils were used for all subjects. To keepwithin a reasonable duration, two viewer locations were used forpresentation and three for testing (Fig. 2). Each combination ofpresentation and testing location was used twice, for a total of 12trials: 4 same-view trials randomly interleaved with 8 shifted-viewtrials. Jon was impaired relative to controls in both conditions. How-ever, he performed well above chance on same-view trials (0° shift),worse after viewpoint shifts of 55°, and at chance after viewpoint shiftsof 85° and 140° (Fig. 6a). A ceiling effect in the controls’ performanceprevents us from quantifying the relative difference in performancebetween the two conditions. This is addressed in the next experiment.

Does Jon have a specific deficit in the shifted-view spatial rec-ognition task over and above his deficit in the same-view task? Totest this we needed to reduce controls’ performance so that theirperformance in the same-view condition was equal to or inferior toJon’s, and then compare performance in the shifted-view condi-tion. Pilot studies indicated that performance did not dependstrongly on the delay between presentation and recall, leaving list-length and number of foils to be varied. We repeated the experi-ment, with Jon tested with two foils and control subjects with 5.List-lengths were also varied, with longer lists being used in theeasier same-view condition (trials of 4, 7, 10, and 13 objects), andshorter lists in the shifted-view condition (trials of 1, 2, 3, 4, 5, and7). Note that list lengths 4 and 7 were used in both conditions topermit direct comparison. To maximize sensitivity only one view-point shift was used (140°; between locations 1 and 3 in Fig. 2).

The second experiment revealed a massive deficit in Jon’s allo-centric spatial memory. The controls no longer performed at ceil-ing on the same-view task but continued to perform slightly worsein the shifted-view condition at comparable list lengths (Fig. 5a).By contrast, Jon’s performance is strikingly different between con-ditions. Tested from the same view, his performance is at ceilingand equal to or better than controls up to 7 items, begins to fall offfor longer lists, and is still above 50% at 13 items (i.e., he has a memoryspan of at least 13). From the shifted view, however, he only performsabove chance when there is a single object to remember (i.e., a memoryspan of 1). Direct comparison between Jon’s performance and that ofthe controls at list lengths of 4 and 7 shows a clear dissociation, withJon performing better than the controls in the same view conditionbut markedly worse in the shifted-view condition (Fig. 5b).

The results indicate that Jon has a selective deficit in shifted-view spatial recognition memory over and above his deficit insame-view spatial recognition memory. This is strong evidence fora deficit in allocentric but not egocentric spatial memory, given

FIGURE 6. a: The effect of varying angle of rotation. Jon (whitebars) shows greater impairment on the shifted-view conditions of the3-D object location task than on the same-view. Ten objects werepresented in the same-view condition and seven in the shifted viewconditions. Testing involved forced-choice recognition with two foilsfor controls and for Jon. Control subjects (black bars) show slightlygreater variation in performance in the shifted-view conditions, butare otherwise unaffected by the view manipulation, possibly due to aceiling effect. Scores are mean percentage correct, error bars are 1 SD.b: Comparison of response latencies as a function of the change inview orientation between presentation and testing, showing the meanand standard error of the median of control subject’s raw responselatencies. Angles refer to the bearing of the subjects view in the envi-ronment (angles of rotation of bearing of view location about thecentroid of the placeholders were 0°, 32°, 62°, 90°, 120°, and 152°).Subjects performed six trials of each rotation, three clockwise andthree anticlockwise, and made three responses per trial. There is amonotonic dependence on angle of rotation (Pearson’s correlation �0.444, P < 0.001), which is also found individually for the first, secondand third responses in each trial (1st response, r � 0.437, P < 0.001; 2ndresponse, r � 0.316, P < 0.001; 3rd response, r � 0.389, P < 0.001).


that performance in the shifted-view condition necessarily reflectsallocentric memory while that in the same-view condition couldreflect either egocentric or allocentric memory.

Does John also have a deficient egocentric spatial memory? Inboth experiments, Jon performed significantly worse than controlson the same-view condition at list lengths 10 and above. Ourinterpretation is that controls can use both egocentric and allocentricsystems, whereas John has only the egocentric one. This suggests thatthe allocentric system can act in parallel to the egocentric one, but canstore more locations or store locations for longer. This may reflect itsability to store a single, amalgamated, representation of the environ-ment and the locations of all the objects presented in a single trial,rather than relying on sensory snapshots of the presentation of eachobject (Burgess et al., 1999; see also Milner et al., 1999; Wang andSpelke, 2000). This would explain Jon’s increasing impairment withincreasing list length, and implication of the hippocampus in spatialtasks that are not necessarily allocentric but involve delays or long lists.This interpretation makes the interesting prediction that there wouldbe no difference in performance between Jon and controls on a pureegocentric task in which allocentric information was of no use.

What processes are involved in solving the shifted-view spatialmemory condition? To investigate this, we looked for evidence of achronometric relationship between the angle or distance between pre-sentation and retrieval positions and the latencies of correct responses.We first examined the response latencies of the twelve control subjectsin experiment 4c. These data did not show a linear relationship be-tween response latencies and view rotation (Pearson’s correlation �0.08, P � 0.64). However, this experiment included only changes inview orientation of 0°, 55°, 85°, and 140°, (corresponding to changesof location of 0, 66, 85, 151 in the bearing of the view location aboutenvironmental centroid, or changes of 0, 36.4, 39.1, and 55.1 virtualmeters), and the 55° and 85° rotations occurred on different sides ofthe environment (between locations 1 and 2, and between locations 2and 3, respectively), and so are confounded by perceptual differences.Accordingly, we performed a further experiment, designed to be moresensitive to trends in response latencies. This experiment involvedtwenty four control subjects, list length 4 (to maximize the numbers ofcorrect responses), two presentation locations (locations 1 and 3; Fig.2) and six testing locations: locations 1 and 3 in Figure 2 and four morelocations spread evenly between them. The results show a clear corre-lation between response latency and the size of the change in vieworientation, both overall and individually for the first, second andthird responses per trial (Fig. 6b).

The response latency data imply that controls’ performance ofthe (allocentric) shifted viewpoint recognition memory task in-volved an imagined movement between the subject’s viewpointrelative to the array of locations. This could be either (1) a move-ment of viewpoint around the array of locations-to-choose-from atretrieval to bring it into correspondence with a stored representa-tion of the object’s location at presentation, (2) a movement ofviewpoint within a stored representation of the object’s location atpresentation to bring this into correspondence with the locations-to-choose-from at retrieval, or (3) a rotation of the array as seenfrom a fixed viewpoint. The first two possibilities would view Jonas having an impaired ability to update locations to reflect animagined movement of viewpoint. In case (1), this deficit would

apply to locations in the perceived world, while in case (2) it wouldapply to remembered locations. The last possibility (3) would viewJon as being incapable of rotating arrays of objects. We note that allof these processes can also be seen in terms of the allocentric andegocentric systems interacting to help reorient the subject to thenew perspective (e.g., Hermer and Spelke, 1994) although this wouldonly predict chronometric response times for the first response in eachtrial. Interestingly Jon was able to perform shifted viewpoint recogni-tion of a single object location successfully. This preserved ability iscompatible with his normal performance on tasks of mental rotation(e.g., 27/32 in the Little Man Test (Ratcliff, 1979), which is shared byother bilateral hippocampal (Holdstock et al., 2000b) and unilateraltemporal lobectomy (Abrahams et al., 1997) patients. It may alsorelate to the fact that, when only a single location is tested, imaginedrotation of an array can be almost as accurate as an equivalent imag-ined movement of viewpoint (Wraga et al., 2000). We conclude thatJon has a preserved ability to rotate single objects from a fixed view-point, but not to perform imagined movements of viewpoint.

In addition to the processes of viewpoint movement, discussedabove, comparison between a stored viewpoint-dependent (i.e.,egocentric or iconic) representation of presentation and the arrayof locations presented at retrieval is required, under conditions inwhich one or other has been subjected to a movement of view-point. Impairment to the processes of storage and comparisonthemselves cannot explain the difference in performance betweensame-view and shifted-view conditions. However, it could be thatJon’s deficit arises from the nature of the stored representationbeing unable to support the 3-D change in viewpoint which requiresknowledge of the objects’ locations. For example this would be true oficonic sensory-bound “snapshot” like representations, which we thinkJon can store successfully (Spiers et al., 2001a). An additional consid-eration is that, since the objects in a trial are presented sequentially,representations of their egocentric locations would each have to bestored and manipulated individually, imposing a high memoryload for long lists. Thus another process likely to aid performancewould be the amalgamation of egocentric representations into asingle enduring representation that is orientated along the view-point at presentation. Again, Jon’s deficit might arise from animpaired ability to form such a representation insofar as such arepresentation was required to support movements of viewpoint.

Thus, the hippocampus appears to support the creation or stor-age of representations of location within which 3-D movements ofviewpoint are possible, or supports the processes of performingthose movements of viewpoint themselves. As such, it provides amechanism that supports allocentric memory: enabling viewpointindependent behavior, despite relying on viewpoint-dependent(i.e., egocentric) representations. Interestingly, Jon may have someinsight into his inability to deal with shifted-viewpoints. In a pilotexperiment in which subjects moved freely between the presenta-tion and retrieval locations, Jon, but not control subjects, stoppedfrequently to look back at how the array altered as he moved (thisstrategy proved unsuccessful for the 7-item lists tested).

Previous studies have implicated the medial temporal lobes (par-ticularly in the right hemisphere) in various tasks requiring mem-ory for object locations (Smith and Milner, 1981; Smith et al.,1995; Abrahams et al., 1997; Bohbot et al., 1998; Abrahams et al.,

818 KING ET AL.

1999 Holdstock et al., 2000b), landmarks (Habib and Sirigu,1987), spatial scenes (Pigott and Milner, 1993) and spatial navi-gation (Maguire et al., 1998). Many of these impairments likelyreflect damage outside the hippocampus, e.g., recognition ofscenes or landmarks (Habib and Sirigu, 1987; Aguirre andD’Esposito, 1997; Epstein and Kanwisher, 1998) and navigationin very simple environments (Bohbot et al., 1998), consistent withJon’s unimpaired recognition of visual scenes (Spiers et al., 2001a).The impairments in the studies involving long delays, large num-bers of locations and movement of the subject likely reflect hip-pocampal damage, consistent with the impairments and interpre-tation presented here. Of these studies, only (Holdstock et al., 2000b)explicitly compared memory for a single location from a shifted view-point with that from a fixed viewpoint. In their study, a patient withhippocampal damage was found to have a greater impairment in theshifted viewpoint condition. Again, this is consistent with our inter-pretation. However, in the study conducted by Holdstock et al., thedifference in performance between the conditions was delay depen-dent (only approaching significance after 20- or 60-s delays) and de-pended in part on the performance of control subjects showing greatervariance in the fixed viewpoint condition (performed in the dark) thanin the shifted viewpoint condition (performed in the light).

Our results are consistent with the proposed role of the hip-pocampus in supporting a cognitive map (O’Keefe and Nadel,1978). In addition, our results indicate that the allocentric behav-ior supported by the hippocampus in this experiment is achievedby mental movements of viewpoint within 3-D space applied toviewpoint-dependent (i.e., egocentric or iconic) representations (assuggested by O’Keefe, 1993). We note that the idea of a “flexiblerelational” memory system (Cohen and Eichenbaum, 1993) en-compasses all of the spatial characteristics of a cognitive map, andmany other nonspatial behaviors (transitive inference, Bunsey andEichenbaum, 1996; social transmission of food preferences, Bun-sey and Eichenbaum, 1995). This theory is thus equally compati-ble with our results, and cannot be distinguished from the cogni-tive map theory on the basis of tests using spatial paradigms.However, perhaps because of its generality, the “flexible relational”idea does not help to provide a mechanistic understanding of thecognitive processes underlying our result. By contrast, the consid-eration of viewpoints within spatial reference frames suggested bythe cognitive map provides the basis for our current interpretationof these processes. In terms of Jon’s nonspatial behavior the flexi-ble-relational hypothesis can also be argued to be consistent withJon’s spared recognition and impaired recall (Baddeley et al.,2001), although this goes against the initially close link betweenrelational memory and “declarative” memory (e.g., Eichenbaum etal., 1996; Squire, 1992) and is less consistent with Jon’s relativelypreserved and fluently expressed semantic knowledge. It is worthnoting also that our results are not simply predicted as a conse-quence of Jon’s acknowledged episodic memory deficit. Retrievingthe event of an object’s presentation corresponds at least as well tothe fixed view condition as to the shifted-view condition. Con-versely, it may be profitable to investigate the deficit in episodicrecollection in terms of the spatial constraints that might apply tothe retrieval of information relating to personally experiencedevents. One possibility is that the hippocampal system constrains

the set of information retrieved or reconstructed from memory tobe consistent with perception from a single position. In this case,where multimodal information is retrieved, the information fromeach modality must be constrained to be consistent with percep-tion from the same single position as applies to the other modali-ties. A second, related, possibility is that the hippocampal systemallows the search through sets of retrieved information consistentwith perception from positions defined by bodily movements fromother, well-remembered, positions (Burgess et al., 2001; Burgess,2002). In this case, the time taken to perform the movement maybe reflected in response times. By focusing on the spatial domainwe have attempted to provide specific mechanistic constraints on asubset of Jon’s impairment. We hope this will provide a step to-ward understanding the processes underlying the more generalrelational, declarative, or episodic characterizations of his impair-ment that are not addressed in the present report.

In summary, we have demonstrated a massive impairment inrecognition of object locations when viewpoint is shifted betweenpresentation and testing compared with when it is not (e.g., mem-ory span falls from at least 13 to 1) in a patient with focal hip-pocampal pathology. Our findings have at least two more generalimplications. First, they provide a well-controlled test that is ex-tremely sensitive to hippocampal pathology. Given Jon’s 50% re-duction in hippocampal volume and much milder mnemonic im-pairment than patients of other etiology (Rempel-Clower et al.,1996; Zola-Morgan et al., 1986) the size of his impairment in ourtask implies that it might detect much milder hippocampal pathol-ogy, such as that shown early in Alzheimer’s disease (e.g., Fox et al.,1996) or preterm children (Isaacs et al., 2000). Second, the veryselective impairments in recognition memory in these patients mayhold important clues for their debilitatingly severe impairmentsacross all tests of recollection of episodic memory. This study pre-sents the most severe impairment yet reported in a recognitionmemory paradigm for these patients. The results hint that episodicrecollection might be related to the movement of viewpoint withina stored representation (see Burgess et al., 2001). This in turnmight explain why the same anatomical systems (mammillary bod-ies, anterior thalamus, subicular/hippocampal complex) are in-volved in both the representation of head-direction (Taube, 1998)and episodic recollection (Aggleton and Brown, 1999).

Acknowledgments

The authors thank Adam Duffen and Will Brookings for col-lecting control data, David Gadian for the imaging of Jon’s brain,Antonio Incisa della Rochetta for useful discussion of the 2-Dobject-location task, and an anonymous referee for directing ustoward the literature on mental rotation.

REFERENCES

Abrahams S, Pickering A, Polkey CE, Morris RG. 1997. Spatial memorydeficits in patients with unilateral damage to the right hippocampalformation. Neuropsychologia 35:11–24.


Abrahams S, Morris RG, Polkey CE, Jarosz JM, Cox TC, Graves M,Pickering A. 1999. Hippocampal involvement in spatial and workingmemory: a structural MRI analysis of patients with unilateral mesialtemporal lobe sclerosis. Brain Cogn 41:39–65.

Aggleton JP, Brown MW. 1999. Episodic memory, amnesia, and thehippocampal–anterior thalamic axis. Behav Brain Sci 22:425–490.

Aguirre GK, D’Esposito M. 1997. Environmental knowledge is subservedby separable dorsal/ventral neural areas. J Neurosci 17:2512–2518.

Baddeley A, Vargha-Khadem F, Mishkin M. 2001. Preserved recognitionin a case of developmental amnesia: implications for the acquisition ofsemantic memory? J Cogn Neurosci 13:357–369.

Bohbot VD, Kalina M, Stepankova K, Spackova N, Petrides M, Nadel L.1998. Spatial memory deficits in patients with lesions to the righthippocampus and to the right parahippocampal cortex. Neuropsycho-logia 36:1217–1238.

Bunsey M, Eichenbaum H. 1995. Selective damage to the hippocampalregion blocks long-term retention of a natural and nonspatial stimu-lus–stimulus association. Hippocampus 5:546–556.

Bunsey M, Eichenbaum H. 1996. Conservation of hippocampal memoryfunction in rats and humans. Nature 379:255–257.

Burgess N. 2002. The hippocampus, space and viewpoints in episodicmemory. Q J Exp Psychol 55:1057–1080.

Burgess N, Jeffery KJ, O’Keefe J. 1999. Integrating hippocampal andparietal functions: a spatial point of view. In: Burgess N, O’Keefe J,Jeffery K, editors. The hippocampal and parietal foundations of spatialcognition. Oxford University Press. p 3–29.

Burgess N, Becker S, King JA, O’Keefe J. 2001. Memory for events andtheir spatial context: models and experiments. Philos Trans R SocLond B Biol Sci 356:1493–1503.

Christou CG, Bulthoff HH. 1999. The perception of spatial layout in avirtual world. Technical report No.(75), Max Planck Institute forBiological Cybernetics, Tubingen, Germany.

Cohen NJ, Eichenbaum H. 1993. Memory, amnesia and the hippocam-pal system. Cambridge, MA: MIT Press.

Diwadkar VA, McNamara TP. 1997. Viewpoint dependence in scenerecognition. Psychol Sci 8:302–307.

Easton RD, Sholl MJ. 1995. Object-array structure, frames of reference, and re-trieval of spatial knowledge. J Exp Psychol Learn Mem Cogn 21:483–500.

Eichenbaum H, Schoenbaum G, Young B, Bunsey M. 1996. Functionalorganization of the hippocampal memory system. Proc Natl Acad SciU S A 93:13500–13507.

Epstein R, Kanwisher N. 1998. A cortical representation of the local visualenvironment. Nature 392:598–601.

Fox NC, Freeborough PA, Rossor MN. 1996. Visualisation and quanti-fication of rates of atrophy in Alzheimer’s disease. Lancet 348:94–97.

Gadian DG, Aicardi J, Watkins KE, Porter DA, Mishkin M, Vargha-Khadem F. 2000. Developmental amnesia associated with early hy-poxic-ischaemic injury. Brain 123:499–507.

Habib M, Sirigu A. 1987. Pure topographical disorientation: a definitionand anatomical basis. Cortex 23:73–85.

Hermer L, Spelke ES. 1994. A geometric process for spatial reorientationin young children. Nature 370:57–59.

Holdstock JS, Gutnikov SA, Gaffan D, Mayes AR. 2000a. Perceptual and mne-monic matching-to-sample in humans: contributions of the hippocampus,perirhinal and other medial temporal lobe cortices. Cortex 36:301–322.

Holdstock JS, Mayes AR, Cezayirli E, Isaac CL, Aggleton JP, Roberts N. 2000b.A comparison of egocentric and allocentric spatial memory in a patientwith selective hippocampal damage. Neuropsychologia 38:410–425.

Isaacs EB, Lucas A, Chong WK, Wood SJ, Johnson CL, Marshall C, Vargha-Khadem F, Gadian DG. 2000. Hippocampal volume and everyday mem-ory in children of very low birth weight. Pediatr Res.47:713–720.

Kinsbourne M, Wood F. Short-term memory processes and the amnesicsyndrome. In Deutsch, JA, editor. Short-term memory. New York:Academic Press. 1975. p 257–291.

Maguire EA, Burgess N, Donnett JG, Frackowiak RS, Frith CD, O’KeefeJ. 1998. Knowing where and getting there: a human navigation net-work. Science 280:921–924.

Maguire EA, Vargha-Khadem F, Mishkin M. 2001. The effects of bilat-eral hippocampal damage on fMRI regional activations and interac-tions during memory retrieval. Brain 124:1156–1170.

Manns JR, Squire LR. 1999. Impaired recognition memory on the Doorsand People Test after damage limited to the hippocampal region.Hippocampus 9:495–499.

McNamara TP, Rump B, Werner S. 2002. Egocentric and geocentricframes of reference in memory of large-scale space. Psychonomic Bul-letin & Review (in press).

Milner AD, Dijkerman HC, Carey DP. 1999. Visuospatial processing ina case of visual form agnosia. In: Burgess N, O’Keefe J, Jeffery K,editors. The hippocampal and parietal foundations of spatial cogni-tion. Oxford: Oxford University Press. p 443–466.

Mou W, McNamara TP. 2002. Intrinsic frames of reference in spatialmemory. J Exp Psychol 28:162–170.

O’Keefe J. 1993. Kant and the sea-horse. In: Eilan N, Brewer B, Mc-Carthy R, editors. Spatial representation: problems in philosophy andpsychology. Oxford: Blackwell. p 43–64.

O’Keefe J, Nadel L. 1978. The hippocampus as a cognitive map. Oxford:Oxford University Press.

Pigott S, Milner B. 1993. Memory for different aspects of complex visualscenes after unilateral temp. Neuropsychologia 31:1–15.

Ratcliff G. 1979. Spatial thought, mental rotation and the right cerebralhemisphere. Neuropsychologia 17:49–54.

Rempel-Clower NL, Zola SM, Squire LR, Amaral DG. 1996. Three casesof enduring memory impairment after bilateral damage limited to thehippocampal formation. J Neurosci 16:5233–5255.

Shelton AL, McNamara TP. 2001. Systems of spatial reference in humanmemory. Cogn Psychol 43:274–310.

Shepherd RN, Metzler J. 1971. Mental rotation of three-dimensionalobjects. Science 171:701–703.

Simons DJ, Wang RF. 1998. Perceiving real-world viewpoint changes.Psychol Sci 9:315–320.

Smith ML, Milner B. 1981. The role of the right hippocampus in therecall of spatial location. Neuropsychologia 19:781–793.

Smith ML, Leonard G, Crane J, Milner B. 1995. The effects of frontal- ortemporal-lobe lesions on susceptibility to interference in spatial mem-ory. Neuropsychologia 33:275–285.

Spiers HJ, Burgess N, Hartley T, Vargha-Khadem F, O’Keefe J. 2001a.Bilateral hippocampal pathology impairs topographical and episodicbut not recognition memory. Hippocampus 11:715–725.

Spiers HJ, Maguire EA, Burgess N. 2001b. Hippocampal amnesia. Neu-rocase 7:357–382.

Squire LR. 1992. Memory and the hippocampus: a synthesis from find-ings with rats, monkeys, and humans. Psychol Rev 99:195–231.

Taube JS. 1998. Head direction cells and the neuropsychological basis fora sense of direction. Prog Neurobiol 55:225–256.

Tulving E. 1983. Elements of episodic memory. Oxford: Clarendon Press.Van Paesschen W, Connelly A, King MD, Jackson GD, Duncan JS. 1997.

The spectrum of hippocampal sclerosis: a quantitative magnetic reso-nance imaging study. Ann Neurol 41:41–51.

Vargha-Khadem F, Gadian DG, Watkins KE, Connelly A, Van PaesschenW, Mishkin M. 1997. Differential effects of early hippocampal pathol-ogy on episodic and semantic memory. Science 277:376–380.

Wang RF, Simons DJ. 1999. Active and passive scene recognition acrossviews. Cognition 70:191–210.

Wang RF, Spelke E. 2000. Updating egocentric representations in humannavigation. Cognition 77:215–250.

Wraga M, Creem SH, Proffitt DR. 2000. Updating displays after imag-ined object and viewer rotations. J Exp Psychol Learn Mem Cogn26:151–168.

Zola-Morgan S, Squire LR, Amaral DG. 1986. Human amnesia and themedial temporal region: enduring memory impairment following abilateral lesion limited to field CA1 of the hippocampus. J Neurosci6:2950–2967.

820 KING ET AL.

king - human hippocampus and spatial memory

Documents